Environmental constraints severely restrict crop yields in most production environments, and expanding the use of variation will underpin future progress in breeding. In semi-arid environments boron toxicity constrains productivity, and genetic improvement is the only effective strategy for addressing the problem1. Wheat breeders have sought and used available genetic diversity from landraces to maintain yield in these environments; however, the identity of the genes at the major tolerance loci was unknown. Here we describe the identification of near-identical, root-specific boron transporter genes underlying the two major-effect quantitative trait loci for boron tolerance in wheat, Bo1 and Bo4 (ref. 2). We show that tolerance to a high concentration of boron is associated with multiple genomic changes including tetraploid introgression, dispersed gene duplication, and variation in gene structure and transcript level. An allelic series was identified from a panel of bread and durum wheat cultivars and landraces originating from diverse agronomic zones. Our results demonstrate that, during selection, breeders have matched functionally different boron tolerance alleles to specific environments. The characterization of boron tolerance in wheat illustrates the power of the new wheat genomic resources to define key adaptive processes that have underpinned crop improvement.
Although wheat is the most important source of calories and protein for much of the world’s population, its large and complex genome has made it recalcitrant to molecular technologies. However, in comparison with model organisms, wheat has the advantages of extensive monitoring and archiving of genotypes and associated phenotypic data and the availability of unique populations adapted to specific environments and end-uses that have resulted from a long history of selective breeding. Early farmers and then modern breeders selected lines adapted to specific environments but, as with most crops, only a small proportion of the available variation in landraces and wild relatives has been effectively captured in breeding programmes3. Understanding the molecular basis for key adaptive traits would provide a powerful strategy for targeting new sources of variation. Rapidly expanding wheat genomic resources are now improving the tractability of molecular studies in wheat. Here we show the power of coupling germplasm collections with genomic resources by investigating wheat adaptation to high concentrations of boron in the soil.
In plants, boron is essential but has a narrow optimal range. Boron toxicity occurs in dry environments, often where plants are grown on alkaline soils of marine or volcanic origin, but sometimes also as a consequence of irrigation4. In bread wheat (Triticum aestivum L.; genomes AABBDD), boron toxicity results in decreases in root growth (Fig. 1a, b), above-ground biomass and yield. Conversely, boron deficiency is associated with high-rainfall climates and soils prone to depletion of mobile elements, resulting in poor seed set or sterility5. In a large study involving 233 trials across Australia over 12 years6, boron-tolerant genotypes had an up to 16% yield advantage over intolerant genotypes in southern wheat-growing regions where boron toxicity has been noted, but a yield disadvantage at sites in northern regions, demonstrating environment-specific adaptation. Quantitative trait loci (QTL) associated with boron tolerance are known in wheat2,7 and barley8, but the only cereal genes that have been shown to have a significant role in tolerance are from barley. In barley the 4H tolerance gene (HvBot1) encodes an anion-permeable transporter that is tandemly duplicated and highly expressed in the tolerant line9, and the 6H tolerance gene (HvNIP2;1) encodes a member of the NIP aquaporin family10. Similarly, in rice and Arabidopsis, genes of these two families have been implicated in responses to boron supply11. However, direct orthologues of these genes do not co-locate with Bo1 or Bo4, the major boron tolerance QTL in wheat. Bo1 is located on chromosome 7BL in the bread wheat cultivar Halberd7,12 and the durum wheat (T. turgidum L. var. durum; genomes AABB) cultivar Lingzhi Baimong Baidamai (abbreviated as Lingzhi)13. Bo4 is located on chromosome 4AL in the bread wheat landrace G61450 (ref. 1).
Bot-D2a (TaBor2, GenBank accession number EU220225; Extended Data Table 1) was previously proposed to be responsible for boron tolerance in bread wheat14, and boron transporter sequences Bot-A4a, Bot-B4a and Bot-D4a (TaBOR1.2, TaBOR 1.3 and TaBOR 1.1) were described recently15. Wheat gene nomenclature guidelines are described in the Supplementary Discussion. Here we show that Bot-D2a maps to chromosome group 3 (Extended Data Fig. 1a), and Bot-A4a, Bot-B4a and Bot-D4a locate to group 5 (International Wheat Genome Sequencing Consortium website, http://www.wheatgenome.org) and are not associated with major loci involved in boron tolerance. We previously reported a fine map of the 7BL region containing Bo1 in the bread wheat doubled haploid population Cranbrook (intolerant) × Halberd16. Here we genotyped 1,700 individuals with markers developed from genes in syntenic regions on Brachypodium supercontig 1 and rice chromosome 6 to identify 153 lines recombinant between barc32 and AWW355. For each F2 recombinant, boron tolerance of F3 progeny families was assessed by root growth in high boron hydroponics. This reduced the target interval containing the tolerance gene to 0.06 centimorgans (Extended Data Fig. 2a). Gene annotation in the syntenic interval from related grass species did not reveal candidate genes for Bo1.
In parallel, we screened BACs derived from the orthologous interval in Aegilops tauschii (genome DD)17 with a probe derived from HvBot1 (ref. 9). A positive clone (HI148P11) was used to identify the sequence of a D-genome HvBot1-like gene. A genomic DNA fragment (AWW461) of this gene co-segregated with Bo1. Sequencing of the gene from Halberd showed that the 7BL tolerance locus contains an undescribed boron transporter-like gene with 80% open reading frame (ORF) similarity to HvBot1 and Bot-D2a. The Halberd gene (Bot-B5b; Fig. 2a) contains 12 introns and encodes a predicted membrane protein of 660 amino-acid residues (GenBank accession number KF148625). We confirmed membrane localization in planta by confocal imaging of onion epidermal cells transiently expressing the construct 35S:Bot-B5b:GFP (Extended Data Fig. 3a), and heterologous expression in Saccharomyces cerevisiae indicated that Bot-B5b is able to function as a boron transporter (Fig. 2c). Using plant boron transporter ORF sequences we constructed the maximum-likelihood phylogeny of this gene family, verifying that Bot-B5 genes have no direct orthologous sequences in barley (Supplementary Discussion) or in the sequenced reference genotypes of Brachypodium distachyon, Oryza sativa (ssp. japonica) or Sorghum bicolor (Extended Data Fig. 4a, b).
Gene similarity between bread wheat Bo1 and Bo4 was investigated by mapping AWW461 in recombinant inbred lines (RILs) derived from a cross G61450 × Kenya Farmer, in which Bo4 was previously linked to XksuG10-4A (ref. 18). We found two copies, one explaining 79% of total trait variation for absolute root length under high boron in the 4AL interval Xabg390-4A–XksuG10-4A, the other locating to the Bo1 locus on 7BL with no significant marker-trait association (Extended Data Fig. 2b). Investigation of the extent of localized similarity around the Halberd 7BL and G61450 4AL genes showed no evidence of chromosomal translocation, indicating that Bo4 represents a dispersed duplication19 of the 7BL gene. Similarly, in durum wheat we mapped a Bot-B5b-derived marker (AWW555-SacI) in F2 plants of Jandaroi (intolerant) × AUS14740 (tolerant) and found co-segregation with AWW5L7, a 7BL-specific marker tightly linked to Bo1 (ref. 20), supporting orthology of the bread and durum wheat 7BL loci. Our findings implicate boron transporters in boron tolerance at three major tolerance loci: Bo1 in bread and durum wheat, and Bo4 in bread wheat. Across 9.4 kilobases (kb) of genomic sequence, G61450 4AL and Lingzhi 7BL genes are identical and differ from Halberd 7BL by a single non-synonymous nucleotide. High sequence identity together with pedigree information (Genetic Resources Information System for Wheat and Triticale, http://www.wheatpedigree.net/) implicates tetraploid lines as sources of Bo1 and Bo4 in hexaploid bread wheat (Supplementary Discussion). Analysis of neighbouring genes on 7BL in current bread wheat cultivars indicates conservation of the tetraploid-derived segment, presenting a barrier to recombination and the exploitation of agronomically important loci linked to Bo1, such as resistance to late-maturity α-amylase21.
A population mutagenized with ethyl methanesulphonate (EMS) was developed in Halberd. At high boron, two independent mutants with intermediate root growth were identified (Extended Data Fig. 5a). Mutant EMS405 contained a substitution in exon 6 of Bot-B5b (Ala 215→Val), which reduced yeast growth under high boron in a complementation assay (Fig. 2c). Mutant EMS388 contained a single nucleotide substitution (G in wild type, A in mutant) within the Bot-B5b gene promoter sequence, 215 base pairs (bp) upstream of the predicted messenger RNA start site. No difference in expression of the Bot-B5b transcript was observed between EMS388 and Halberd. Both mutants co-segregated with reduced root growth under high boron (Extended Data Fig. 5b).
Bread and durum wheat Bot-B5 alleles broadly fell into three groups. The first comprised all boron-tolerant cultivars containing the Halberd (Bot-B5b), G61450 (Bot(Tp4A)-B5c) or Lingzhi (Bot-B5c) alleles. Intolerant lines fell into two further allele groups: those related to the gene in the reference genotype Chinese Spring (Bot-B5a), and those in which the gene was partly or wholly deleted (Fig. 2a). In Cranbrook (hexaploid) and Langdon (tetraploid), which contain the null allele Bot(Df)-B5h (Extended Data Fig. 1b), the genomic deletion is estimated at more than 22 kb and includes the complete Bot-B5 gene. Chinese Spring-group alleles (Bot-B5a, Bot-B5d and Bot-B5e) have 98% ORF sequence identity to Bot-B5b and are characterized by the insertion of repetitive sequences in the promoter region 2,240 bp upstream of the mRNA start codon site, in addition to protein sequence polymorphism compared with Halberd Bot-B5b (Fig. 2a, b). Protein function comparison in yeast under high boron showed reduced function of Bot-B5a compared with both Bot-B5b and Bot-B5d (Fig. 2c). Bot-B5a and Bot-B5d differ by only two residues, suggesting key roles of one or both residues. Alleles in the Halberd and Chinese Spring groups both showed root-specific expression that is responsive to high boron (Extended Data Fig. 3b, c), but they differed in transcript level (Fig. 2d), consistent with variation in observed root length phenotypes under high boron (Fig. 1a). The low level of expression found for the Chinese Spring group may have been due to the insertions in the promoter sequence. High expression of Bot-B5 in G61450 was derived from the 4AL allele Bot(Tp4A)-B5c, illustrated by comparison between G61450 × Kenya Farmer-derived lines RIL26 and RIL30 (Fig. 2d). Bot-B5c in durum wheat and Bot(Tp4A)-B5c in G61450 showed similar expression levels, consistent with sequence identity and recent transposition of a functional gene.
No homoeologous 7A sequences have been found in bread wheat, in durum wheat or in A genome progenitor species, supporting the observation20 that boron tolerance is absent among T. monococcum and T. urartu accessions, and implying an Aegilops source for this boron transporter in cultivated wheat. In Chinese Spring (see http://www.wheatgenome.org), we found a D genome sequence (Bot-D5a) low in expression that contained a 4-bp frame-shift mutation encoding a truncated, non-functional protein of 56 residues (Fig. 2a, b). We originally identified Bot-B5b through the bridging genotype Ae. tauschii accession AL8/78, so examined the synthetic hexaploid wheat SW58 derived from Langdon (intolerant) × AL8/78. We identified a transcript (Bot-D5b) more highly expressed than Bot-D5a with 97% protein identity to Halberd Bot-B5b (Fig. 2a, b). In a yeast complementation assay, Bot-D5b was functional but had lower efficacy than Bot-B5b, supporting plant root length data that demonstrate less tolerance in SW58 than Halberd, but greater tolerance than Langdon (Extended Data Fig. 6a, b).
In total we identified 12 sequence variants for the Bot-B5/D5 genes, and we can account for boron tolerance phenotype on the basis of gene sequence: boron intolerance results from a loss of function through complete or partial deletion, or frame shift mutation (Cranbrook, AUS10110b, Kenya Farmer and Chinese Spring Bot-D5a), partial loss of function or reduced effect from induced mutation (EMS405 and EMS388), and decreased transcript level (Chinese Spring Bot-B5a, G61450 Bot-B5d and AUS30656 Bot-B5e). Three highly conserved but distinct natural variants (Halberd, Lingzhi and G61450 Bot(Tp4A)-B5c) are fully functional.
The adaptive advantages of different boron tolerance alleles was demonstrated by genotyping a set of 85 released cultivars and 153 advanced breeding lines, revealing a biased deployment of Bot-B5 alleles between southern and northern Australian wheat-growing regions (Fig. 3): tolerance alleles predominated in southern regions and were absent in lines targeted to northern regions. Despite relatively early introduction into Australian breeding programmes, the highly expressed Bot(Tp4A)-B5c allele was not detected in lines from either region. Furthermore, genotyping of boron-tolerant lines from diverse locations outside Australia failed to identify lines carrying both Bot-B5b and Bot(Tp4A)-B5c, consistent with a penalty associated with the presence of strong tolerance alleles in low-boron environments.
The challenge in expanding the variation available to breeders depends on the identification and subsequent deployment of novel variation. Selection of wheat lines adapted to different production environments has been occurring since wheat was domesticated about 10,000 years ago; in selecting for performance, early farmers around the Mediterranean, through the Middle East and into northern India, Afghanistan and China developed wheat landraces with varying levels of tolerance to high concentrations of boron in the soil (Fig. 3). A similar process was occurring in barley, but our data reveal that boron tolerance in barley and wheat arose through the divergent evolution of paralogous genes (Extended Data Fig. 4a, b). In wheat the generation of comparatively broad allelic variation provided adaptation to agro-geographically diverse regions. Plant breeders in the early twentieth century recognized the value of landraces as sources of useful variation and exploited locally adapted lines, leading to the emergence of boron-tolerant varieties in the Mediterranean. Wheat is a relatively recent introduction into Australia and the Americas, with early varieties based on lines adapted to temperate, usually northern European, environments. Problems of adaptation in these new environments led breeders back to landraces, where diversity was sought in accessions derived from regions with perceived environmental similarities (Fig. 3). In this study we have shown that for both bread and durum wheat this empirical approach was successful and resulted in the release of boron-tolerant cultivars in Australia and South America. We have also shown that there has been strong selection for or against functionally different Bot-B5 alleles in contrasting environments, which implies that matching boron tolerance alleles to the level of soil boron is critical in achieving maximum yield potential.
Phenotyping for boron toxicity response
The length of the longest root of 8–14-day-old seedlings grown either in aerated hydroponics or on moistened filter paper was measured as described18,20 except that in the hydroponics assays the low-boron treatment was 0.015 mM boron.
Genetic manipulations followed established protocols. Sequence data were obtained from 454 sequencing of Chinese Spring BAC clones 112M01 and 451O08 (ref. 22), from Sanger sequencing of genomic and cDNA fragments and from databases of wheat genomic sequences23. Gene expression analysis of allelic variants used four biological replicates comprising root tissue from seedlings grown hydroponically for 16 days in 0.05 mM boron followed by 22 h in 2 mM boron.
Functional assessment in yeast
Yeast strain INVSc2 was used in all experiments. Growth experiments in liquid medium and on solid medium were as described previously9. Data from time-course growth assays at low and high boron concentrations were plotted and fitted to Boltzmann sigmoidal functions by using nonlinear modelling (GraphPad Prism 6) to calculate times to half-saturation for five replicates.
Transient expression in onion epidermal cells
The Bot-B5b coding sequence was cloned into vector pMDC83 to generate a carboxy terminus construct, 35S:Bot-B5b:GFP, and introduced into onion epidermis by bombardment. Cells were visualized by confocal image analysis before and after plasmolysis.
Complete ORF sequences were aligned using MUSCLE, and the maximum likelihood based on the Kimura 2 parameter model was calculated. Bootstrap values were generated from 1,000 replicates.
Mapping and QTL analyses were performed with MapManager QTX version 0.30 (ref. 24).
Plant and DNA materials
Bread wheat lines used for phenotyping assays and whole-gene sequencing were Cranbrook, Chinese Spring, G61450, Halberd, Kenya Farmer, Wl*MMC, and lines from a G61450 × Kenya Farmer recombinant inbred line (RIL) mapping population of 90 lines18 (RIL26, RIL30, and tolerant and intolerant G61450 x Kenya Farmer RIL pools). RIL26 and RIL-4AL pool (RIL6, RIL7, RIL9, RIL26, RIL62, RIL66, RIL87 and RIL91) contain a G61450 allele at Bo4 on chromosome 4AL (functional allele Bot(Tp4A)-B5c) and a Kenya Farmer allele at Bo1 on chromosome 7BL (truncated allele Bot-B5g). RIL30 and RIL-7BL pool (RIL16, RIL21, RIL27, RIL30, RIL44, RIL56, RIL93 and RIL97) have a Kenya Farmer allele at Bo4 on 4AL (null) and a G61450 allele at Bo1 on 7BL (functional allele Bot-B5d). Durum wheat lines used for phenotyping assays and whole-gene sequencing were Langdon and the landraces AUS10110 (Uttar Pradesh, India), AUS10344 (Triticum durum Desf. var. niloticum, Iraq), AUS14010 (Lingzhi Baimong Baidamai, China) and AUS14740 (Afghanistan). AUS10110, AUS10344 and AUS14010 were previously identified as boron-tolerant in a screen13 of 300 genotypes from North Africa, Asia, Australia, Italy and the International Maize and Wheat Improvement Center (CIMMYT). In our study, AUS10110 was identified as segregating for boron tolerance and found to be heterogeneous at the 7B Bot-B5 locus. Selections of AUS10110 were made based on allele composition, with AUS10110a containing the functional allele Bot-B5c and AUS10110b containing the truncated and non-functional allele Bot-B5f. Synthetic wheats used for phenotyping assays, genetic analysis and whole-gene sequencing were SW58 (Langdon × Ae. tauschii AL8/78) and AUS30656 (LCK59.61/Ae. tauschii). SW58 was supplied by S. Wu.
Cultivars, breeders’ lines or DNA samples for marker screening were obtained from various Australian wheat breeding programmes, D. Mares and the Australian Winter Cereals Collection (AWCC). DNA of Turkish cultivars was supplied by T. Oz. DNA of 112 F2 lines from the tetraploid population Jandaroi × AUS14740 was supplied by N. Shamaya.
Halberd-EMS mutagenized lines were generated by treatment of wheat cv. Halberd seed with 0.45% (v/v) ethyl methanesulphonate (EMS) for 16 h. Putative mutant EMS-Halberd M3 seedlings were identified by phenotyping about 15–20 seeds per M2 family in high-boron hydroponics (10 mM boron) in a greenhouse for 10–14 days. Boron-intolerant families were identified on the basis of short root length and the appearance of the first leaf (tip necrosis, yellow wilting). Individual putative mutant plants were selected and transplanted into hydroponics without boron for recovery. After approximately 2 weeks of recovery hydroponics, survivors were transplanted into pots of soil for seed multiplication. Further phenotype validation undertaken in M4 and M5 generations and full gene sequencing of M4 pools resulted in the identification of two Bot-B5b mutant families, EMS388 and EMS405.
Hydroponic phenotyping for boron toxicity response
Seeds for phenotyping assays, with the exception of the G61450 × Kenya Farmer RIL population, were germinated on filter paper and grown hydroponically in a greenhouse or controlled-environment chamber (20 °C day/15 °C night, 12-h day) for 1 day in a low-boron minimal nutrient base solution containing 0.5 mM Ca(NO3)2, 2.5 μM ZnSO4 and 0.015 mM boron as H3BO3 and then for 8–14 days, with aeration, in either fresh base solution (low-boron treatment) or base solution plus additional boron (4–10 mM boron) as H3BO3. Solutions were replaced once or twice, depending on the length of treatment. Seedlings showing fungal infection around the seed were discarded. Absolute length of the longest root (RL) was measured for 13–16 seedlings per line at each treatment, and average RL and s.e.m. were calculated. Previous studies have shown that RL under high boron is highly correlated with boron tolerance12,20. Phenotyping of the G61450 × Kenya Farmer RIL population was performed similarly, except that the seedlings were grown in 9.3 mM boron (100 mg kg−1 boron) on filter paper as described previously27.
Plants for quantitative RT–PCR (qRT–PCR) and northern analysis were grown in two experiments (experiments 1 and 2). In both experiments, plants were grown for 17 days in a full nutrient base hydroponics solution containing 5 mM NH4NO3, 5 mM KNO3, 2 mM Ca(NO3)2, 2 mM MgSO4, 0.1 mM KH2PO4, 0.05 mM NaFe(iii)EDTA, 0.05 mM H3BO3, 5 μM MnCl2, 10 μM ZnSO4, 0.5 μM CuSO4 and 0.1 μM Na2MoO4 with aeration, in a controlled-environment growth room at 22 °C (day)/16 °C (night) with a 14-h photoperiod. Solutions were replaced every 3–4 days during the experiment. Seeds were germinated on filter paper, and seedlings with shoots of 2–3 cm were transplanted to hydroponics. The experiment design was four biological replicates per genotype for each of three treatments, and seedlings within a treatment were arranged in a modified Latin-square pattern. In experiment 1, each biological replicate comprised a pool of two seedlings to reduce sampling error further. In experiment 2, each biological replicate comprised a single seedling. Treatments were low boron for 17 days, 22 h at 2 mM boron in full nutrient solution applied at day 16, and 7 days at 2 mM boron in full nutrient solution applied at day 10. All treatments were harvested on day 17.
Mapping and QTL analysis
Marker linkage in the Jandaroi × AUS14740 F2 population, and marker linkage and single marker regression QTL analyses in the G61450 × Kenya Farmer RIL population, were all performed with MapManager QTX version 0.30 (ref. 24). Genetic map images were generated with MapChart 2.2 software28.
Nucleic acid extraction, Southern and northern analysis, rapid amplification of cDNA ends (RACE), cDNA synthesis and quantitative real-time PCR
Genomic DNA was extracted using either of two standard methods29,30; phenol/chloroform-extracted DNA was used for Southern analysis. Total RNA was extracted from roots of hydroponically grown plants with TRIzol (Invitrogen) followed by ISOLATE plant RNA spin column purification (Bioline). We synthesized first-strand cDNA using Superscript III reverse transcriptase (Invitrogen) and used it as the template to amplify Bot-B5 transcripts. qRT–PCR assays were performed with methods described previously31. SMART RACE (Clontech) cDNA synthesis was used to obtain cDNA for determining 5′ and 3′ mRNA sequences of Halberd and Chinese Spring Bot-B5 transcripts. Southern and northern analysis using 32P-labelled probes was performed with standard methods. Final washing after probe hybridization for both Southern and northern membranes was in 0.5 × SSC, 0.1% SDS solution for 20 min at 65 °C.
For Southern analysis to locate Bot-D2a in wheat (Extended Data Fig. 1a), we digested genomic DNA from Chinese Spring nullisomic–tetrasomic (CS N-T) chromosome substitution lines with DraI and hybridized with the 32P-labelled probe AWW469, a 261-bp cDNA fragment of Bot-D2a amplified from Cranbrook, which does not cross-hybridize to Bot-B5 or Bot-D5 at high stringency. We included a genotype of the D-genome species Ae. tauschii to assist in interpretation. For Southern analysis to demonstrate both the absence of 7B Bot-B5 sequences in Cranbrook and Langdon, and the absence of homoeologous 7A sequences in bread and tetraploid wheat (Extended Data Fig. 1b), we used the bread wheat cultivars Cranbrook, Halberd, Chinese Spring and CS N7B-T7A, and the reference durum wheat cultivar Langdon. Genomic DNA was digested with HindIII and hybridized with the 32P-labelled probe AWW471, a 357-bp genomic DNA fragment of Bot-D5b that hybridizes to Bot-B5 at high stringency.
For northern analysis of Bot-B5 transcript induction under high-boron conditions (Extended Data Fig. 3c) we used root tissue from Halberd grown in two independent experiments (experiment 1 and experiment 2, described in detail above) and G61450 seedlings grown in one of the experiments (experiment 1). To increase replication in experiment 2, where each biological replicate comprised a single plant, two sets of Halberd lines were sampled. Hybridization with a 268-bp cDNA probe (AWW548) derived from Halberd, comprising 65 bp of coding sequence and 203 bp of 3′ UTR, was used to detect Bot-B5 transcripts.
Semi-quantitative RT–PCR was performed on a Chinese Spring developmental tissue series32 using the Bot-B5-specific marker qRT–PCR-Bot-B5 (36 cycles) and a wheat glyceraldehyde-3-phosphate dehydrogenase (GAPDH) marker (28 cycles). PCR products were separated on 1.5% agarose in 1 × Tris-acetate-EDTA buffer and revealed with ethidium bromide by standard methods.
Allele diversity and markers for Bot-B5 and Bot-D5
Sequence data for allelic variants of the Bot-B5 and Bot-D5 genes were derived from fragments amplified from both genomic DNA and cDNA templates produced from individual genotypes. PCR was generally performed using Immolase DNA polymerase (Bioline), with 36 cycles of amplification. PCR products were purified using NucleoSpin II (Macherey-Nagel) or ISOLATE (Bioline) kits. Sanger sequencing was performed using BigDye V3.1 (ABI). Sequences were aligned using Contig Express software (Vector NTI Advance 11.0, Invitrogen). Gene sequences are available under the accession numbers listed in Extended Data Table 1.
Sequencing of Chinese Spring 7B BAC clones 112M01 and 451O08 (ref. 22), in addition to accessing a database of Chinese Spring genomic sequences33, yielded a 27.3-kb contig containing Bot-B5. Contig authenticity was verified by PCR from a Chinese Spring genomic DNA template of fragments spanning joined sequence blocks. We confirmed sequence accuracy for a 9.4-kb genomic region spanning Bot-B5a by Sanger sequencing of PCR products amplified from Chinese Spring and a Chinese Spring nullisomic 7D–tetrasomic 7B template. In the same way we generated a corresponding sequence across the 9.4-kb genomic region for selected other lines. For each of these lines a transcript sequence was also obtained to verify coding region sequences and to assess variation in splice form. In all lines containing full-length Bot-B5 alleles, a region of intron 4 contained a string of G residues about 20 bp in length that prevented through-sequencing. Sizing of fragments on 3% agarose gels, in combination with sequence data from both strands, was consistent with sequences containing only the G-string. Sequences for Bot-D5a and Bot-D5b alleles were obtained from root cDNA of Chinese Spring and SW58, respectively and were verified against genomic sequence databases and Sanger sequencing of genomic fragments.
A suite of allele-discriminating PCR markers were developed for determining the presence of Bot-B5 alleles a–h, with the exception of Bot-B5e. Bot-B5e was detected in a Mexican-derived synthetic wheat line and was therefore considered unlikely to be a common allele either globally in bread wheat or in Australian tetraploid germplasm. Primer sequences are provided in Supplementary Table 1. The dominant marker AWW525, which detects only Bot-B5b, Bot(Tp4A)-B5c and Bot-B5c, is based on a 24-bp duplication in the 3′ UTR region, 206 bp from the TGA stop codon, and is not found in other alleles. To further distinguish Bot-B5b from Bot(Tp4A)-B5c and Bot-B5c, a CAPS marker, AWW532-HpyAV, based on the single exon-3 single nucleotide polymorphism was used. Co-dominant marker AWW600, based on promoter sequence differences, was used to distinguish Bot-B5a from Bot-B5d/Bot-B5e alleles. Lines having other allele types yield no AWW600 product. Lines carrying the truncated allele Bot-B5g or the null allele Bot(Df)-B5h were identified in a two-step process. The first step used the dominant marker AWW555 to identify null types, because it yields a product for all alleles except Bot(Df)-B5h. The second step used the dominant marker AWW516, which is located in the region of Bot-B5 deleted in the Bot-B5g allele, spanning intron 4, and yielding products of different sizes from 7B and 7D genomes. Lines carrying Bot-B5g or Bot(Df)-B5h alleles yield only a single product, derived from the D-genome, whereas lines carrying full-length Bot-B5 alleles yield two products. The use of AWW516 in conjunction with AWW555 overcame the possibility of an incorrect assignment of a null allele in the instance of a failed PCR reaction. Allele data were determined for 153 Australian advanced breeding lines, 85 Australian cultivars and 54 non-Australian hexaploid and tetraploid lines (Fig. 3).
Functional assessment of Bot-B5 and Bot-D5 alleles in yeast
Full-length coding sequences of each of Bot-B5a, Bot-B5b, Bot-B5d, Bot-B5b-EMS405 and Bot-D5b were cloned in the Gateway entry vector pCR8 (Invitrogen). Inserts were confirmed by sequencing and transferred to a Gateway-enabled destination vector for yeast expression, pYES-DEST52 (Invitrogen).
Yeast (Saccharomyces cerevisiae strain INVSc2; Invitrogen) were transformed using a standard lithium acetate method34. Growth experiments on solid medium were conducted as described previously9 to compare the boron tolerance of Bot-B5b-expressing clones and Bot-D5b-expressing clones with each other and with that of yeast transformed with a truncated non-functional version of the Chinese Spring Bot-B5a allele (Bot-B5a-sv). Boron tolerance of Bot-B5-expressing clones was quantified by culturing yeast in minimal liquid medium containing 2% galactose as a source of carbon, both at low boron and with an additional 15 mM H3BO3. Growth was recorded by removing aliquots of cell suspensions at intervals and measuring the attenuance (D600) with a spectrophotometer.
Phylogenetic analysis of plant boron transporter genes sequences
DNA sequences were either identified in this study or obtained from gene sequence databases23,35,36 and the International Wheat Genome Sequencing Consortium (http://www.wheatgenome.org). Sequences were trimmed to cover the complete ORF. Triticum uratu and Ae. tauschii sequences are 99–100% identical to orthologous bread wheat genes and were not included in the phylogenetic analysis, with the exception of AetBot-D5b. Similarly, sequences of Bot-B5c and Bot(Tp4A)-B5c are nearly identical to that of Bot-B5b and not included. Nomenclature is in accordance with internationally accepted guidelines for wheat gene nomenclature and symbolization37. The rice locus LOC_Os01g08020 contains two boron transporters of 96% sequence identity but is currently annotated as a single gene comprising a combination of the two transporters. Using transcript support from DQ421408 and AK072421 we generated two putative ORF sequences designated LOC_Os01g08020_gene A and LOC_Os01g08020_gene B. Phylogenetic analysis was performed using MEGA5 (ref. 38). Sequences were aligned using MUSCLE, and the maximum likelihood based on the Kimura 2 parameter model was calculated. To model evolutionary rates among sites, a discrete gamma distribution (two categories) was used (Γ = 1.1336). The tree with the highest log likelihood (−18,743.4333) is shown in Extended Data Fig. 4a. Bootstrap values were generated from 1,000 replicates.
Transient expression in onion epidermal cells
Bot-B5b coding sequence was cloned into the vector pMDC83 to generate a 35S:Bot-B5b:GFP (C terminus) construct. The 35S:Bot-B5b:GFP construct was used to transform onion (Allium cepa) epidermal cells by particle bombardment, and cells were visualized by confocal image analysis, as described previously39. Plasmolysis of onion epidermal cells was performed by immersion in 1 M sucrose for 1 min before image analysis.
Unless otherwise described, data were analysed using one-way analysis of variance (ANOVA) and Tukey’s multiple comparisons tests (α = 0.05) for differences between treatments, using GraphPad Prism 6 software. qRT–PCR data of Bot-B5 transcript levels (Fig. 2d) were log10-transformed before analysis. For analysis of root growth of genotypes across different boron concentrations shown in Fig. 1a, data for each genotype between 4 and 10 mM boron were compared by linear regression and Tukey’s post-hoc testing for significant differences between slopes. For analysis of yeast growth (Fig. 2c), optical density data were plotted and fitted to Boltzmann sigmoidal functions by using nonlinear modelling with GraphPad Prism, to calculate times to half-saturation. Times to half-saturation (n = 5) were then compared by one-way ANOVA and Tukey’s Honestly Significant Difference test.
We thank N. Collins, J. Dvorjak, H. Kuchel, D. Mares, S. Wu, the John Innes Centre, the Biotechnology and Biological Sciences Research Council, the Institut National de la Recherche Agronomique and the International Wheat Genome Sequencing Consortium for resources, and J. Tiong, T. Oz and A. Pohlen for assistance. The authors are supported by grants from the Australian Research Council, the Grains Research and Development Corporation and the South Australian Government.
Extended data figures
Extended data tables
This table displays PCR markers and probes for genotyping and expression analysis. a, Primer sequences and expected amplicon sizes for PCR markers derived from Bot-B5/D5 which were used for allele typing and Bot-B5 expression analysis. b, primer sequences for probe generation and c, primer sequences for PCR markers used for F2 recombinant screening and semi-quantitative RT-PCR analysis.
This table shows wild and cultivated barleys screened for TaBot-B5b orthologues. Name and source of genotypes screened by Southern analysis with AWW548.